Forms of Bias Flashcards
What is Bias?
Inclination of scores in a certain (usually favourable) direction, not just random error.
Statistical regression
Random component to the severity of symptoms in chronic conditions. Chronic conditions have a cyclical pattern.
Pt’s seek care when symptoms are more severe, more likely next fluctuation will be towards average symptom level.
Placebo effect
Improvement in outcomes as a result of treatment ritual
Non-specific treatment effects
Hawthorne effect
Changes in behaviour because its known behaviour is under scrutiny. Patients may over or under exaggerate symptoms.
Rosenthal effect
Theapist led - high expectation lead to improved performance
Polite patient effect
Patient may be reluctant to disappoint therapist and exaggerate recovery when they in fact aren’t feeling better.
Recall bias
Therapists more readily remember their successes - extreme cases feature more prominent in memory.
Ways of controlling bias for natural history or statistical regression
- use of comparison group (group that doesn’t get the treatment)
- groups need to be 100% equivalent
- randomisation to avoid selection bias
- subtract the change due to natural history/other treatment
Ways of controlling bias for placebo, Rosenthal and Hawthorne
- sham intervention
- blind patient: ritual but no Rx
- blind therapist: don’t know which treatment they are giving (hard to do in physio), the ritual has to be the same
Ways of controlling bias for polite patient effect and detection bias
- independent assessors: removed from clinical process
- assessors blind to group allocation
- sometimes the patient is the assessor*
How to control for Selection bias
- comparable groups -> randomisation
- must have a control group
- adequate sequence generation
- concealed randomisation/allocation
How to control for Attrition bias
- keep drop out low (less than 15%)
- keep group random: state where, when & amount lost
- sensitivity analysis can help decide if biased
- use Intention to treat analysis
Definition and How to control for Performance bias
participants or therapists desire for a study to fall a certain way
- blinding of patients (unknown group allocation)
- blinding of therapists (unaware of treatment allocation - equal enthusiasm)
How to control for Reporting bias
- non-selective report
- state primary outcomes
- report primary outcomes appropriately
- between-group analysis
- point estimate and variability
- primary outcomes should relate to the focus of the study
What is spin?
Reporting practices that distort the interpretation of results and mislead readers so that results are viewed in a more favorable light.